silver fir–beech forests in the western Pyrenees
2.3.1. Species classification and general data exploration
Vascular plant species were categorized following Schmidt et al. (2011) and classified based of the information contained in the Vegetation-Plot Database of the University of the Basque Country, BIOVEG (EU-00-011; Biurrun et al., 2012) and of the available reference list about beech forest specialists (Willner et al., 2009). Two ecological groups were distinguished: forest specialists (FS), species largely restricted to closed forests, and non-forest species (NF) which included forest edge species, generalists and open area species.
Bryophytes were classified according to their life-strategy type (Dierssen, 2001) into short-lived (SL) and long-short-lived (LL) species. The category short-short-lived includes fugitives, annual shut-tles, colonists, ephemeral colonists, pioneer colonists, short-lived shuttles and geophytes, and the category long-lived includes long-lived shuttles, competitive perennials, stress tolerant perennials and dominant perennials. All observed bryophytes were identified to species level, except some specimens of the genus Ulota: only specimens with a developed capsule were identified to species level, otherwise (in 11 plots) they were identified to genus level only.
Spe-Table 1. Environmental variables available for 32 plots used in the analyses of plant species diversity in the western Pyrenean silver fir–beech forests. Differences between managed and unmanaged stands were tested using Mann-Whitney’s U test. ∗ p < 0.05, ∗∗ p < 0.01 and ∗∗∗ p < 0.001.
Managed Unmanaged
Variables Unit Mean±SD Mean±SD p value
Topography
Elevation m a.s.l. 1118±104.70 1447±140.76 5.096e-06***
Slope % 46.62±16.21 39.19±12.40 0.19
Light conditions
Transmission % 10.81±6.13 11.20±2.67 0.36
Gaps % 6.7±5.50 5.9±2.28 0.89
LAI 4.28±0.96 4.04±0.79 0.54
Climate
(T) mean annual temperature ºC 9.46±0.32 7.95±0.49 1.07e-06***
(P) annual rainfall (mm) mm 1422±73.39 1384±72.33 0.28
(It) thermicity index 137.6±11.03 89.73±13.62 1.383e-06***
(Ic) continentality index 15.95±0.42 16.51±0.40 0.0002***
(Io) ombrothermic index 12.63±0.98 14.15±1.10 0.0015***
Soil
pH (1:2,5 V/V) 5.4±0.37 5.7±0.33 0.02*
Coarse sand (0.2-2.0 mm) % 2.35±0.95 4.95±2.42 0.002**
Fine sand (0.002-0.2 mm) % 21.95±6.14 29.63±8.08 0.006**
Clay (< 0.002 mm) % 32.91±3.55 29.76±5.85 0.05
Silt (0.002-0.02 mm) % 42.80±4.91 35.94±5.06 0.0003***
Organic matter % 7.62±2.11 7.94±2.01 0.61
Nitrogen Kjeldahl % 0.27±0.03 0.63±0.11 0.01*
Phosphorus mg/l 3.31±0.94 3.96±1.69 0.38
Potassium mg/l 88.31±25.22 69.31±16.09 0.004**
Calcium mg/l 1459.8±1135.09 1336.4±484.2 0.73
Sodium mg/l 8.43±2.85 6.18±2.42 0.02*
Magnesium mg/l 101.62±44.93 86.69±32.12 0.27
Aluminium saturation (Al%) % 15.01±20.66 5.26±10.46 0.06
Stand structure
Tree density (N/ha) 638±286 806±617 0.83
Basal area of living trees (m2/ha) 34.95±9.41 35.20±12.61 0.79
Basal area snags (m2/ha) 1.21±2.47 2.27±2.75 0.06
Basal area logs (m2/ha) 3.89±3.91 2.21±1.80 0.46
Total basal area of dead wood (m2/ha) 5.09±4.18 4.53±2.77 0.95
Number of large trees (N/ha) 84.29±29.61 88.77±41.85 0.83
Number of large fir trees (N/ha) 39.49±35.20 32.80±41.85 0.35 Number of large beech trees (N/ha) 45.29±31.67 55.97±33.85 0.35
VarD 50.40±18.23 62.36±17.77 0.06
* CEC - cation exchange capacity
cies-level identifications (Ulota crispa, U. bruchii and U. crispula) were used in the general spe-cies lists, but for the quantitative numerical analyses we used genus-level identifications (Ulota sp.)
Differences between the two management categories in stand structure, climatic conditions and soil properties were assessed by the Mann-Whitney’s U test.
2.3.2. Species richness
Generalized linear mixed models were used to determine which environmental variables best explain variation in species richness of all plants, vascular plants and bryophytes, as well as of the ecologically defined subgroups (FS, NF, SL and LL). Environmental variables were taken as fixed effects, and stand, which was nested within management type, was considered as a ran-dom effect. Poisson error distribution for count data and log link function were used. Environ-mental variables were standardized to zero mean and unit variance, and Spearman correlation analysis was performed among them to detect collinearity. Drivers of species richness were assessed using multimodel inference. The full model and all possible sub-models were fitted according to Information-Theoretic (IT) approach (Burnham and Anderson, 2002) using dredge function in MuMIn package in R. Models were ranked based on their AICc, which is a criterion used as a measure of model quality for small sample sizes. The best fitting sub-models, i.e.
those within a distance of 2 AICc units from the first ranking model, were tested for overdis-persion. We also tested whether the residuals of any model showed significant spatial auto-correlation using the ape package in R, but they did not (results not shown). To assess the magnitude of a possible management effect, we calculated the log response ratio, a metric that expresses proportional differences in species richness between managed and unmanaged stands. Effect sizes of other variables were displayed using theeffects package in R.
2.3.3. Species turnover
To assess which factors explain variation in species turnover, we opted for a distance-based variation partitioning approach. This was done both because pairwise dissimilarity indices provide a natural measure for species turnover, and because a distance-based approach allows directly incorporating geographical distances into the models. By variation partitioning, we quantified the contribution of environmental differences and geographical distances to ex-plaining species turnover, which helps to separate between the effects of dispersal limitation and of species responses to spatially autocorrelated environmental variables (Tuomisto and Ruokolainen, 2005; Lichstein, 2007; Sabatini et al., 2014). The response variable, species tur-nover, was calculated using the Sørensen dissimilarity measure (calculated as the one-com-plement of the Sørensen similarity index that expresses the number of shared species as a proportion of the mean number of species in the plots being compared). Separate analyses were run for species turnover of all plants, vascular plants and bryophytes. The explanatory variables were dissimilarity matrices calculated for each environmental variable separately using the Euclidean distance, and a matrix of straight-line geographical distances. For each plant group, we first ran a set of simple Mantel tests (999 permutations) using each explanatory variable separately. All the variables that returned a significant Mantel test result were then used in Multiple Regression on Distance Matrices (MRM), which was simplified by backward elimination to retain only those explanatory variables that had a significant partial contribution to the final model. The contribution of each variable on its own and in combination with the other variables to explaining variation in species turnover was then quantified by running a series of MRM models using different subsets of the retained explanatory variables.
The differences in species composition among plots were visualised, for each plant group se-parately, with NMDS ordination (non-metric multidimensional scaling) based on the Sørensen dissimilarity matrices. To find a stable solution, 500 NMDS ordinations with random starting configurations were run in each case. The environmental variables retained in the final MRM model were then passively projected on the corresponding ordination diagram. Finally, the correspondence between vascular plant and bryophyte species turnovers was quantified using a Mantel test. Multivariate analyses were performed with vegan and ecodist packages in R. All analyses were run using (v. 3.2-5) R statistical environment (R Foundation for Statistical Com-puting, Vienna, AT).
3. RESULTS